Skip to content
AgentOpen-source agent

DeepAnalyze

Open 8B agentic model that runs a full data science pipeline from raw data

Pricing
Open source
Platforms
self-hosted, Web
GitHub stars
4.8K
Last commit
15 days ago
Language
Python
License
MIT

About DeepAnalyze

DeepAnalyze is an open-source agent for analytics and data. It is available on self-hosted and Web.

DeepAnalyze is a research project from a university data lab. Its centrepiece is DeepAnalyze-8B, a language model trained to handle data science work by itself rather than following a fixed workflow.

Given databases or files such as CSV, Excel, JSON, XML, YAML, TXT and Markdown, it prepares data, analyses it, builds models, draws charts and writes up a research report. The authors describe a curriculum-based training method and a trajectory synthesis framework, and report that the 8B model beats workflow-based agents built on proprietary LLMs.

The model, code, training data and demo are all released under MIT, so teams can deploy it or adapt it into their own analysis assistant.

What it does

End-to-end pipeline
Covers data preparation, analysis, modeling, visualization and report generation.
Open-ended research
Produces analyst-style reports from varied data sources.
Many input formats
Reads databases, CSV, Excel, JSON, XML, YAML, TXT and Markdown.
Curriculum training
Trained progressively from single abilities to complex real-world tasks.
Fully open
Model, code, training data and demo are released.

Good for

  • Automate exploratory analysis on a dataset
  • Generate a research report from mixed data files
  • Self-host a data analysis assistant
  • Study agentic training methods

Pricing

Open source

Free, MIT-licensed.

Prices change. Check ruc-deepanalyze.github.io before you buy.

DeepAnalyze FAQ

How large is the model?

The paper introduces DeepAnalyze-8B, an 8-billion-parameter model.

What data can it use?

Databases, CSV, Excel, JSON, XML, YAML, TXT and Markdown.

Is it open source?

Yes, the repository is MIT-licensed and the model, code and training data are open.

Is DeepAnalyze free?

Yes. DeepAnalyze is open source, so you can use and self-host it for free.

What are the best alternatives to DeepAnalyze?

The closest alternatives to DeepAnalyze are deer-flow, DeepCode, daily_stock_analysis, ai-agent-book and Vibe-Trading.

Alternatives to DeepAnalyze

  1. Open-source super agent harness that researches, codes and creates with sandboxes

    • Open source
    • Open-source agent
    • 85.4K GitHub stars
    • AI models and LLM tooling
  2. Autonomous framework that turns research papers into working code repositories

    • Open source
    • Open-source agent
    • 16.8K GitHub stars
    • Coding
  3. LLM-driven stock analysis producing daily reports for A-share, HK and US markets

    • Open source
    • Open-source agent
    • 72.9K GitHub stars
    • Finance and payments
  4. Open-source book and 109 experiments explaining AI agent design and engineering

    • Open source
    • Hosted agent
    • 58.4K GitHub stars
    • Knowledge and memory
  5. Open-source finance research agent for backtests, swarm analysis and trade journals

    • Open source
    • Open-source agent
    • 38.4K GitHub stars
    • Finance and payments
  6. Multi-agent public opinion analysis system that mines social media and writes reports

    • Open source
    • Open-source agent
    • 42.7K GitHub stars
    • Analytics and data